exa-search

What This Skill Does

The exa-search skill provides a powerful, high-performance interface for the Exa (formerly Metaphor) search engine, specifically designed to integrate into the OpenClaw environment. Unlike traditional search engines, Exa is optimized for AI agents, delivering structured, clean, and highly relevant data points including titles, URLs, snippets, and—optionally—the full content of web pages. By leveraging this skill, you can perform deep-web research, gather up-to-date information, and retrieve structured context for your agent workflows directly from your local terminal using Node.js.

Installation

To integrate this skill into your OpenClaw setup, use the following command in your terminal:

clawhub install openclaw/skills/skills/xinhai-ai/exa-search

Once installed, you must configure your credentials. The skill relies on an API key from Exa. You can set this globally by adding EXA_API_KEY=your_key_here to your ~/.openclaw/.env file or by defining it directly in your Gateway environment. Ensure that your Node.js environment is configured to execute scripts from the OpenClaw base directory.

Use Cases

This skill is perfect for scenarios requiring high-fidelity web retrieval. Use it for competitive analysis, where you need to extract specific text content from competitor landing pages. It is also excellent for research tasks that require a specific temporal focus, such as finding news articles or technical documentation released within a specific date range. Because the output is structured, it serves as an ideal data ingestion tool for downstream AI analysis or content synthesis tasks.

Example Prompts

1. "Exa, find the latest developments in AI agent frameworks from January 2025 to February 2026 and summarize the key trends."
2. "Perform an Exa search for 'best practices in TypeScript configuration' and include the full page text for the top 3 results."
3. "Search for the documentation of the OpenClaw API, limiting the results to 5 items."

Tips & Limitations

• Cost Efficiency: Requesting full page text (--text) consumes more API credits than standard snippet searches. Use this flag only when necessary.
• Temporal Filtering: Utilize the --start and --end flags to filter noise and focus strictly on relevant, recent information.
• Scope: This skill is an alternative to the standard web_search and does not replace it. It should be invoked when the specialized indexing of Exa is required for complex agentic research.